Samuel Boivin | Nurphoto | Getty Photos
If one factor turned clear on the HumanX convention in San Francisco this week, the place 6,500 executives, founders and traders gathered to speak about synthetic intelligence, it is that OpenAI not dominates the dialog of their business. For now, at the least, that distinction belongs to Anthropic.
Anthropic’s viral coding agent, Claude Code, was the software on everybody’s lips, at the same time as many attendees acknowledged that OpenAI, Cursor and Google are providing sturdy alternate options.
Regardless of its spat with the Pentagon that went public final month and shortly made its approach to the courtroom, Anthropic has solely gained momentum. The Division of Protection blacklisted Claude, however after opposing rulings in two courts, Anthropic can preserve working with different federal companies whereas the instances play out.
Anthropic’s early power within the enterprise has positioned it to learn from the hovering reputation of AI coding brokers, that are used to generate, edit and assessment code. So whereas OpenAI kicked off the generative AI increase with the launch of ChatGPT in 2022, Anthropic could also be greatest set as much as win contracts from the most important spenders.
CNBC spoke with 19 executives and traders at HumanX, a few of whom requested to not be named with a view to communicate freely. Listed here are the highest three takeaways.
Claude has ‘grow to be a faith’
Anthropic was based in 2021 by a gaggle of researchers and executives who defected from OpenAI. The startup is valued at $380 billion, making it one of the priceless non-public firms on the planet.
Claude Code launched to most people in Might 2025, and as of February was producing greater than $2.5 billion in annualized income. Arvind Jain, CEO of enterprise AI firm Glean, mentioned Claude Code has impressed “Claude Mania,” which is placing strain on enterprise leaders to deploy it.
“It has grow to be a faith, that is the extent of that mania,” Jain mentioned in an interview. “All people, in the event you go and ask them at present, ‘Hey, if I gave you one AI software, what software would you need?’ The reply could be Claude.”
On Tuesday, Anthropic introduced a brand new AI mannequin, Claude Mythos Preview, with superior cybersecurity capabilities because of its sturdy coding and reasoning abilities. The mannequin sparked lots of buzz at HumanX, despite the fact that its rollout is restricted to a choose group of roughly 50 firms.
Victor Riparbelli, CEO of AI video firm Synthesia, mentioned Anthropic has managed to show focus and restraint with its fashions and product, which could be troublesome for a younger hyper-growth firm.
“The blokes at Anthropic have been identical to, ‘We’re not going to do something about video, we’re not going to care about voice fashions, we’re simply going to resolve code gen,’ and now we’re right here,” Riparbelli mentioned in an interview. “OpenAI has had the issue of getting to market six completely different merchandise, which simply takes up thoughts area for the buyer.”
One investor cautioned that whereas Anthropic has been constant and managed to determine a sticky AI use case, the business continues to be younger, and momentum may simply swing in one other route.
AI change administration
As tech firms work to usher their clients into the AI period, they’re additionally grappling with find out how to leverage and deploy brokers internally. Even for Silicon Valley startups, maintaining with the tempo of change isn’t any simple feat.
Ashwin Sreenivas, president of AI startup Decagon, mentioned the appearance of coding brokers has led to a variety of shifts inside his firm. Decagon has modified its interview course of to permit candidates to make use of the instruments, and the corporate is ready to depend on smaller groups of engineers.
A venture which will have required 4 or 5 engineers “turns into two engineers as a result of everybody can transfer quite a bit quicker and go quite a bit farther,” Sreenivas mentioned in an interview.
For Navrina Singh, CEO of AI governance startup Credo AI, the proliferation of latest AI instruments has been concurrently thrilling and anxiousness inducing for her. Overcommunicating, notably together with her clients, has grow to be important, she mentioned.
“The issues that I couldn’t do final 12 months and I wanted to rent 10 folks, I can truly construct over a weekend and deploy for myself and for the corporate,” Singh mentioned. “The anxiousness is I am unable to management my roadmap, and I am unable to management my commitments to the enterprise clients who love extra readability and who like a bit of bit extra stability.”
Huge tech incumbents are navigating comparable adjustments.
Cisco President Jeetu Patel mentioned roughly 85% of his firm’s engineering workforce, or about 18,000 workers, are utilizing AI, however the path to getting there was in contrast to what he’d anticipated. Patel mentioned Cisco initially realized it i wanted to prioritize adoption over outcomes, and to belief that mannequin capabilities will proceed to enhance.
“You’ll be able to’t consider these as instruments, you need to consider these as digital coworkers which might be becoming a member of your crew, as a result of your composition of your scrum crew adjustments,” Patel mentioned on the convention. “You may not have a scrum crew of eight folks. You might need a scrum crew of two folks and 6 brokers, or two folks and infinite brokers.”
The race in opposition to China
Qwen3 is Alibaba’s newest massive language mannequin, which it says combines conventional LLM capabilities with “superior, dynamic reasoning.”
Sopa Photos | Lightrocket | Getty Photos
The fragile two-week ceasefire settlement between the U.S. and Iran has huge implications for power and monetary markets throughout the globe. However the overwhelming majority of execs and traders who spoke to CNBC at HumanX this week mentioned they don’t seem to be but experiencing any direct enterprise impression from the newest battle within the Center East.
Slightly, they’re targeted on one other looming geopolitical downside: China’s open-weight fashions.
In AI, a mannequin is taken into account open weight if its parameters, or the weather that enhance its outputs and predictions throughout coaching, are publicly obtainable. As of April, Chinese language open-weight fashions, together with GLM-5.1, Kimi K2.5 and Qwen3.5, dominate business benchmarks.
American firms are swarming to China’s fashions. Cursor constructed its Composer 2 mannequin utilizing Kimi 2.5. Airbnb CEO Brian Chesky instructed CNBC in October that his firm’s chatbot was largely depending on Alibaba’s Qwen.
Given the significance the U.S. AI business is putting on beating China in relation to innovation, there is a huge emphasis domestically on closing the hole in open weight. Two traders instructed CNBC they’re dedicating lots of their time and sources to that effort, and a 3rd mentioned it is one of many key issues for the business to resolve proper now.
Glean’s Jain mentioned having a number of choices is important.
“The development that we see is that enterprises at present, they’re very cautious of relying on one or two suppliers for all of their AI,” Jain mentioned. “They do not wish to work with only one mannequin firm, as a result of they know that innovation is going on throughout many and in addition in open supply. You wish to have a selection.”
